Using scaling fluctuation analysis to determine the accuracy of climate records and multiproxy reconstructions and the return periods of climate events

English Abstract:

There has been much debate about the reliability of climate records and multiproxy reconstructions. There are now half a dozen global scale climate records and a dozen or so multiproxy reconstructions, each based on somewhat different data and different methodologies. We argue that a) we can only have high levels of confidence when there is substantive agreement between them, b) due to numerous sources of bias, that the most difficult problem is accurate rendition of the low frequencies. Using three instrumental records and a reanalysis and nine last millenium multiproxy reconstructions, we therefore determined the variabilities as functions of time scale as well as the differences between series as functions of time scale using fluctuation analysis (based on easy to understand and interpret Haar wavelets).

Although due to increased temporal averaging, we might naively expect the discrepancies between records to diminish with scale, on the contrary we find that for the global temperature over the range 1 month to one hundred years, that the RMS fluctuations in the differences are constant at around ±0.03K. While this is sufficient for many purposes, it cannot be easily accounted for - we discuss various hypotheses. An analogous analysis of the multiproxies shows a) an improvement of the more recent ones (post 2003) with respect to the older ones, b) agreement to within about ±0.05 K for scales below a century, but rapidly rising at longer time scales. We conclude that for centennial scale studies of pre-industrial statistics (including the determination of return times for natural temperate fluctuations and statistical hypothesis testing), that the data and multiproxies are sufficiently accurate. However, this may not be true of the multicentennial, and longer scale variability.